THE REFLECTING POWER OF VARIOUS METALS

By W. W. Coblentz

CONTENTSPage

Introduction 197

Tungsten 200

Molybdenum 205

Graphite 206

Tantalum 208

Antimony 211

Rhodium 213

Iridium 213

Iron 2 13

Magnesium 213

Chromium 214

Vanadium 215

Tellurium 215

Silicon 216

Summary 218

Note I.—Thermoelectric Properties of Molybdenum 220

Note II.—Preservation of Silver Mirrors 221

Note III.—Comments on Radiation Laws of Metals 221

INTRODUCTION

The present observations are the result of an attempt to obtain

a more thorough knowledge of the optical properties of the metals

used in incandescent lamps. Several other elements on the border

line between the metals and nonmetals, not heretofore examined,

are included, since the data are of general interest.

In a paper on the radiation constants of metals * attention was

called to the fact that all the pure metals thus far examined have

^his Bulletin 5, p. 339; 1908. So also this Bulletin, 2, p. 470; 1907, for illustra-

tion of the reflectivities of various metals.

197

198 Bulletin of the Bureau of Standards [Vol. 7, No. 2

the common property of a low reflecting power in the ultra-violet

and in the visible spectrum, which rises more or less abruptly to

very high values beyond 2/x in the infra-red, and that it was

therefore not unreasonable to assume that the unexamined metals,

such as tungsten and osmium, have this same property. This

follows from the fact that all the members of a group of chem-

ically related substances have similar physical properties. As

will be noticed on a subsequent page, this assumption of a low

reflectivity in the visible spectrum has been verified to a degree

far beyond expectation.

Since the reflectivity of metals is a smooth and continuous

function of the wave length in the infra-red, there is reason for

expecting the reflectivity curves of these heretofore unexamined

metals (tungsten, tantalum, etc.) also to be smooth and contin-

uous. Their spectral energy curves must therefore be smooth

and continuous in the infra-red (no indentations or protuberances)

,

and experimental curves showing such indentations are to be

regarded with suspicion as to the accuracy of the observations.

It is difficult to show experimentally this selective emission in

other than colored metals (e. g., gold and copper) by means of

emission spectra, and since the emissivity in the short wavelengths appears to have only a small temperature coefficient it is

possible to gain some knowledge of the emissivity of the metal at

high temperatures by determining its reflectivity at low tempera-

tures. That it is permissible to apply data obtained at low tem-

peratures to the same substance operated at high temperatures

follows from the recent work of Hagen and Rubens, 2 in which it

is shown that for wave lengths less than 4/x the optical constants

of metals vary but slightly, if at all, with change in temperature,

which is the region of the spectrum of interest on the question of

the selective emission of incandescent lamps with metal filaments.

We have therefore a positive qualitative proof of the much dis-

cussed and much overworked application of " selective emission"

(combined with a high operating temperature) to explain the high

luminous efficiency of metal filament lamps. As will be noticed

in the appended illustrations, the selective emission must lie in

2 Hagen & Rubens, Phys. Zs., 11, p. 139, 1910; Verh. Phys. Gesell.,10, p. 710; 1908.

Cobientz] The Reflecting Power of Various Metals 199

the visible spectrum. Here the emissivity is almost 50% (for tung-

sten) while in the infra-red it is only from 4 to 10% that of anideal radiator, or " black body."

In the present experiments a fluorite prism, a mirror spectro-

meter, and a new vacuum bolometer were used. The unknownmetal was compared with a new silvered glass mirror of which the

reflecting power is known. The absolute value of the reflectivity

of the unknown metal was obtained by multiplying the observed

reflectivity by the reflectivity of silver given in Table II (data fromHagen and Rubens).

A Nernst glower was used as a source of radiation, whichpermitted making observations in the extreme blue and at 10/x

where the fluorite prism is almost opaque. Since the glower is

very narrow it is necessary to have a very accurate adjustment

of its images, reflected from the two mirrors upon the spectro-

meter slit, so as to avoid a systematic error in the reflectivity. Theobservations in the infra-red were further verified by using a

Nernst "heater" instead of the glower.

In the blue and beyond 8// in the infra-red the accuracy is of

the order of 2 to 3%, while in the intervening region the accuracy

is of the order of 1 per cent. There is no difficulty in obtaining

high accuracy since we are concerned only with ratios of two

numbers (galvanometer readings) which are of nearly the samevalue, and which are subject to the same errors of observation.

In view of the physical nature of the material, and especially of

the size of the specimens obtainable, and also of the indirect

comparison with silver, it would be illusory to attempt to attain

a higher accuracy for the absolute values. This is especially

true of tungsten and molybdenum in which the mirror surface

was ground upon one face of a rectangular prism of the pure

metal 20 x 6 x 6 mm on an edge, kindly prepared for us by the

General Electric Company. The narrowness of the specimen

does not permit the grinding and polishing of an optically plane

surface. The dimensions of the tantalum mirror were 15 x 10 x

1 mm; it was prepared by Siemens and Halske and was optically

plane and highly polished, 'fhe close agreement between the

observed reflectivity of tungsten and molybdenum at 12/x and

the theoretical value obtained from a knowledge of the electrical

200 Bulletin of the Bureau of Standards [Vol. 7, No. 2

conductivity is- therefore to be regarded as somewhat accidental.

The angle of incidence of the light upon the mirrors was from 1

2

to 15 degrees, which is sufficiently small to be considered normal

incidence, and hence the loss by absorption (100-reflectivity) is a

measure of the emissivity of the metal.

From his observations on the optical constants (the refractive

index n, and extinction coefficient k), at 0.579//. Wartenberg 3

computed the reflecting power of various metals, some of which

are included in the present paper. It will be noticed in Table I

that the results obtained by the two methods are in excellent

agreement in all samples having a good polish.

TABLE I

Substance Reflecting Power Substance Reflecting Power

Manganese

Chromium

63.5 p. c.

69.7; *55

57.5; *57.5

43.8; *45

41.3

22.3; *23

Tungsten

Palladium

Rhodium

48.6; *51

65.0

VanadinTn , , 78.3; *77.5

Tantalum ... ...... Platinum 72.5

Niobium Iridium 74.6; *75

Graphite Silicon 35.7; *32

Observations by W. W. C.

It will be noticed that even in the visible spectrum the samemagnitude of the reflectivity is to be found within a group of

metals. For example, in the platinum group the reflectivity does

not fall below 70 per cent, while in the chromiun group the reflec-

tivity centers about the 60 per cent point. However, only beyond12/z is this relationship true in general.

TUNGSTEN

The area of the mirror examined was 5 mm x 17 mm. This

specimen was made from the pure material, which is somewhatbrittle and did not permit a polished surface perfectly free from

pores, except in the center. This causes a slight scattering for

wave lengths up to 3/-1 (see discussion under graphite)

.

3 Wartenberg, Verh. Deutsch. Phys. Gesell., 12, p. 105; 1910. Hennig. Zcit. Instk

30, 61, 1910; gives similar data on the reflectivity observed by direct measurement.

Coblentz] The Reflecting Power of Various Metals

TABLE II

20I

%i0)0<— o

Reflecting Power

S>

35 5

a3

o

K

a5tobfl

a3H

i

4)

Sg>»3

aoH

3iJ Mq,83

aoH

O

a3

OS

>

>>ao

1<

aoo

4)

bo 3

a3*C3"5)

H

9

3COc

•40/i

.46

.5

.6

.7

.8

1.0

1.2

1.4

1.6

2.0

2.5

3.0

3.5

4.0

5.0

6.0

7.0

8.0

9.0

10.0

12.0

12.0 ft

84.0

89.0

90.0

92.5

94.2

95.1

96.4

97.2

97.6

97.8

97.9

97.9

98.0

98.3

98.4

98.5

98.6

98.6

98.8

98.9

99.0

98.9

48

55

58.4

64.2

69.0

70.3

73.0

75.0

76.8

78.0

80.6

88.8

76

79

81

84

86.5

88.3

89.5

91

91.6

92

47.0

48.2

49.3

51.3

54.0

56.3

62.3

68.2

73.8

78.0

84.6

89.2

90.5

92.0

92.8

94.0

•94.6

95.1

95.6

95.5

95.5

96.3

*97.2

44.0

44.6

45.5

47.6

49.8

52.3

58.2

63.6

69.0

74.2

81.6

85.5

87.6

89.2

90.5

92.0

93.0

93.3

93.7

94.0

94.5

95.2

*97.3

21

22

55.0

57.5

59.5

61.5

65.0

68.5

71.5

74.3

78.0

81.5

84.5

87.5

89.5

91.5

93.0

94.0

94.0

93.8

38.0

45

56

64.5

78.5

84

86.5

88

90.5

91.6

92.3

92.8

93.0

93.0

93.5

93.8

55 57.0

53

55

34

32

30

29.2

28

72.0

73.0 49

22.5

23.5

56

57

59

61

63

66

70

76

81

85

89

92

93

58.5

59.5

61.3

63

64.5

66

69

71.7

74.3

76.7

78.8

82

85

87.8

89.8

92

24.0

74.0

48

49.5

50

51

52

53

54

55

57

60

63

68

72

78

25.0

26.8

28.3

58

60

28 75 30.0

32.0

28 77

79

80.5

35.2

39.5

65 28 43.0

46.0

91.5

93.5

95.5

95.1

95.4

95.9

96.5

92.5

93

93.5

93.5

94

94.5

68

70

28

28

83.5

86

88

91

93

93

47.5

50.5

52.0

53.5

55.5

72 28 57.5

59.0

95.0

*95.3 *97r""

* Values computed from the electrical conductivity.

In Fig. i and in Table II it will be noticed that the reflectivity

of tungsten rises abruptly from a low value of 50 per cent in the

yellow to 89 per cent at 2.5/u, beyond which point it increases

gradually to 96 per cent at 10fi. This is a characteristic of pure

metals.

Hagen and Rubens 4 have shown that for long waves (greater

than 1 2ft) the absorption (100-reflectivity) of a metal may be

computed from its electrical conductivity by means of the formula

100 —R = ^7==,where R is the observed reflecting power, c is the

reciprocal of the resistance, in ohms, of a conductor 1 m long and

1 sq. mm in cross section, and X = wave length in /x = .ooi mm.

4 Hagen & Rubens, Ann. der Phys. (4) U, p. 873; 1903. Coblentz, this Bulletin, 2,

p. 470; 1907.

202 Bulletin of the Bureau of Standards [Vol. 7, No. 2

The specific resistance of a squirted rod (hence probably not

so pure as the mirror material) of tungsten 20 cm long and 1.2 mmin diameter was determined by Mr. J. H. Dellinger to be 0.000 006 9

ohm per centimeter cube at 2i°.8 or 3.99 times that of copper.

The sample of tungsten upon which the reflectivity oberservations

were made had a specific resistance of 0.0000063 at 2i°.5 C, or

3.66 times that of copper. Its density was 10 (true value about

19 or 20), which indicates that there were hollow spaces along the

axis of prism of metal, which was about 2.5 cm long and had been

broken from a longer rod. At one end of this specimen there was

a small cavity, but it was not possible to tell how far it extended

90

80

a 7 °

60

50

40

& WoUp— -

Moe- °tr-»

s^^

9u

• 9^/Ad

a

> m

i3 4 5 6 7

TUNGSTEN, MOLYBDENUM

Fig. 1

10/A

along the central axis of the specimen. This would cause a higher

resistance than the true value. Recent determinations by Fink

and by Parani indicate a specific resistance of about 0.000,005

for annealed tungsten. From the specific resistance of this sample

of tungsten the computed reflectivity, from the aforesaid formula,

using X=i2/x, is 97.3 per cent, while the value obtained by extra-

polating from 1 oft is 96.3 per cent. As already stated, this is in

closer agreement than is to be expected from the observations and

from the size of the mirror employed. It may be noticed in

Table II that the infra-red reflectivity of tungsten is the same as

that of platinum, and in the visible it is the same as that of steel.

Coblentz] The Reflecting Power of Various Metals 203

For a perfectly polished surface, free from pores, the reflectivity

of tungsten would be somewhat higher out to 2/x, because of the

elimination of the diffuse reflection which obtained in the present

sample. (See graphite, which gives an extreme case of diffuse

reflectivity.) The results show that in the previous investigation

70

60

50

>1-

>to 40</>

5Id

30

b

20

^a

10 £u O-> cca |_

CO /

n§§m.0 1.0 1.4 1.8

TUNGSTEN

Fig. 2

2.2 2.6 3.0yx

of the radiation constants of metals, it was permissible to assume

that the same emissivity function obtains as in platinum, with, of

course, different numerical values of the constants.

In Fig. 2, curve a, is given the spectral energy curve of a new

1 10-volt, 32-candlepower tungsten lamp when on a normal opera-

71132 —115

204 Bulletin of the Bureau of Standards [Vol. 7, No. 2

tion of 55 watts. The mean value of the radiation constant for

this lamp was previously computed 1 to be about ^ = 6.5, the

variations in the individual computations being rather large.

The spectrometer calibration has just been recomputed, which is

more accurate than the one then employed. The new computation

of this radiation constant is not sufficiently different from the old

one to warrant a revision of the data then published, although the

individual computations are in closer agreement. The new value

of the \moa! = 1.225// (old \max= 1.257/i) and the new a = 6.2, the

shift in the new calibration curve being about 0.02/1 toward the

short wave lengths. On the basis of the new calibration curve

the wave lengths of maximum emission are too large by about

0.02 p. The radiation constants are also too large by 3 per cent

to 5 per cent. Since the individual computations vary by this

amount, and since different samples of the metal filaments seem

to be subject to variation, the old values are probably as trust-

worthy as would be the revised ones. The experimental data

does not indicate a variation of the constant "a" with wavelength. For, at any temperature, the values of "a" are all high

or all low irrespective of the wave lengths selected in the compu-tations.

The reflection curve and the radiation constant {a = 6.2) are in

close agreement with that of platinum. We are therefore in a

position to make comparisons with platinum of which the radiation

constants are fairly accurately known. Using the XmaxT = 262o

of platinum, and the observed ^moa . = 1.225// °f tungsten, the

operating temperature of the aforesaid lamp was 1870 C. Fery

and Cheneveau (Compt. Rend., 149, 777, 1909) by using radiation

methods obtained 1875 .

From the observed radiation curve a, Fig. 2, we can obtain the

black body curve, b, at the same temperature, by dividing the

observed emissivities by the observed absorptivities (100-reflec-

tivity) of tungsten given in Fig. 1. The maximum emission lies

at about Xmax = 1.45/1, from which the computed temperature is

about 1760 C. This value is no doubt somewhat too low, because

no account was taken of the loss by diffusion in the reflection

curve at 0.6 to 2//. Using the reflectivity values of platinum

which are in all probability higher than the true values of tungsten,

Cobientz) The Reflecting Power of Various Metals 205

the XTOaa.= 1.38ft in place of ?Wc = i-45/* and t = i86o° C, which

is probably too high. The computed values of a, from curve b,

Fig. 2, vary from 4.6 to 5.5 while the experimental value for a

"black-body" is a = 5. This shows that curve b is not a true

"black-body" radiation curve. There are two evident reasons

for this discrepancy. First, the observed curve a has not been

corrected for absorption by the glass walls of the lamp, which

would give an emissivity at 2.5/1 somewhat as shown by dotted

lines in Fig. 2. Second, the reflectivity is not known with suffi-

cient accuracy at 1.5 to 2ji (because of diffuse reflection due to

the fine pores in the specimen examined) where a small error in

the reflectivity has a very large effect in the position of the maxi-

mum of the radiation curve. If it were not for this difficulty, this

method could be used for estimating temperatures of metal fila-

ments of known reflecting power (using perfect mirrors for deter-

mining the reflectivity) , but of which it is impossible to measure

the temperature directly. Since the emissivity and the reflectivity

are functions of the refractive index and of the absorption coeffi-

cient, and since in considering the emissivity we are not concerned

with the scattering effect, it does not seem necessary to consider

the question of the "blackening" of the radiation due to the

porosity of the filament, so long as the size of the pores are of the

same magnitude as the wave length of the emitted light. This

case seems to be different from the experiment in which the

mechanical scratches, made in a strip of incandescent metal, show

a higher emissivity than the outer surface. The present observa-

tions and those of Waidner and Burgess 5 seem to indicate that

on "normal" operation the tungsten filament is at a temperature

of 1850 C to 2200 C.

MOLYBDENUM

The area of the mirror examined was 5 mm x 17 mm. The

reflecting surface had a very much higher polish than the sample

of tungsten just described. To the eye it was apparent that the

molybdenum had a higher absorption than the tungsten mirror.

This is illustrated in the bolometric comparison, shown in Fig. 1,

which indicates a reflectivity of only 46 per cent in the yellow,

6 Waidner and Burgess, this Bulletin, 2, p. 319.

206 Bulletin of the Bureau of Standards [Vol. 7, No. 2

which rises abruptly to 85 per cent at 2.5/1 beyond which point

the reflectivity gradually increases to 95 per cent at 10/*. The

specific resistance of two samples of molybdenum wire, drawn to

a uniform diameter, was determined by Mr. Dellinger. The wires

were respectively 15 cm and 20 cm in length, 0.627 mm and 0.815

mm in diameter, and the corresponding specific resistance was

.000 006 4 and .000 006 2 at 21°.^. This is about 3.6 times the

resistance of copper and practically the same as that of tungsten.

From this it appears that the reflectivity of molybdenum should

be as high as that of tungsten at 12/x. The observations indicate

the opposite condition. No systematic errors could be detected

in the observations and the only explanation which can be offered

at present is that the size of the reflecting surfaces was too small

to establish a true absorption to a greater accuracy than 2 per

cent at 12/x. This, however, is not a large discrepancy when it

is recalled that in some of the earlier work, using larger mirrors of

known curvature, errors almost as large as these are on record.

The reflectivity curves of molybdenum and tungsten are so

nearly alike that from a consideration of their emissivities and

luminous efficiencies there seems to be no great choice in the

use of these two metals in incandescent lamps. On the other

hand, from a consideration of their physical properties, the

molybdenum filament would be the preferable because it seems to

retain its toughness and its ductility, in contrast with tungsten

which seems to become crystalline and brittle after being oper-

ated, for a while, at a high temperature. The melting point of

molybdenum is quite high, and it is principally a question of

overcoming certain physical weaknesses, such as disintegration,

in order to make this material useful for incandescent lamp

filaments.

GRAPHITE

Two samples of graphite were examined, the one being the

natural mineral from Siberia. Its reflecting surface was 4 cm x

5 cm. The second sample, curve b%Fig 3, was a longitudinal

section of a rod, 8 mm in diameter, of Acheson's graphite. It

could not be burnished free from pores which were rather large,

and which caused a scattering of the radiation, out to 4/i in the

Coblentz] The Reflecting Power of Various Metals 207

infra-red. Beyond this point it reflects as highly as the natural

S berian mineral, shown in curve a, Fig. 3. The natural mineral

was very compact, but showed the rays and fibrous structure of

the original plant or tree from which it was formed. It wasburnished or "polished" on a semi-matte piece of plate glass. Athoroughly ground glass was too rough, while the smooth plate

was found equally inapplicable. The partly ground glass seemedto hold just sufficient material to work into the softer parts without

accumulating sufficient loose material which would roughen the

60#r-#r—"S •Ta

^s-^-c

50

> 40h-OU

15 30ce

20

cy<?>r

V*

//

10

C

1 2 34 567 8 91 0//,

GRAPHITE, TANTALUM

Fig. 3

surface. The glass plate was kept wet with clean water and bythe application of some pressure it was possible to prepare a very

highly polished surface

The reflectivity is 23 per cent in the yellow, gradually increasing

to 63 per cent at io^t. From the low reflectivity at 2 ft in the

infra-red there results a high emissivity (in contrast with the

metals), hence the graphatized carbon filament can not have the

same luminous efficiency as the tungsten lamp at the same tem-

perature, although it has the higher emissivity in the visible

spectrum.

208 Bulletin of the Bureau of Standards [Vol. 7, No. 2

TANTALUM

The utilized area of the tantalum mirror was 9 mm x 14 mm.In spite of its excellent polish, it showed an exceedingly low

reflectivity throughout the spectrum, as compared with pure

metals.

The curve c, in Fig. 3 represents three series of observations,

using different adjustments, the measurements usually agreeing

to one part in 100 to 200. In the visible spectrum the reflectivity

is only 10 per cent, which gradually increases to 63 per cent at 10/x.

There is no indication of an approach to constant reflectivity, in

in the infra-red, such as obtains in pure metals. Subsequent

correspondence with the maker showed that this surmise of an

impure metal was correct. The surface was rendered impure byworking and polishing which, by the action of gases, formed an

exceedingly hard surface layer. From the writer's experience it

appears that the ease with which a tantalum mirror becomes con-

taminated in polishing depends upon the previous history of the

metal. For example, the regulus of tantalum obtained from

Von Wartenberg 6 had never been rolled or hammered, and it was

found that, in polishing it, any contamination, such as for exam-

ple, the blackening caused by a film of oil, was easily removed.

On the other hand, a film of oil on the mirror just described,

which had evidently been worked down from a regulus (Fig. 3, c)

formed a brown coating which was removed with great difficulty. 7

The rear side of this mirror was therefore ground flat on fine

emery paper, which for the final polish was covered with a layer

of alcohol and graphite. By stroking lightly and lifting the

mirror soon after the alcohol had entirely evaporated, a clean

surface was produced which was almost free from scratches. Theold surface, of which the reflectivity is given in Fig. 3, was so hard

that it could not be polished by this method.

The contrast between the reflectivity of the old contaminated

surface and the new surface, polished as just described, is shown

6 Von Wartenberg, Verh. Phys. Gesell., 12, p. 121; 1910. He found that tantalum

was not contaminated by polishing.7 See further a recent investigation by Parani, Verh. Phys. Gesell., 12, p. 301, 1910,

who likewise finds that tantalum is easily contaminated by gases.

Coblentz] The Reflecting Power of Various Metals 209

in Fig. 4. Curve c gives the reflectivity of the new surface as

produced by the fine dry emery paper, the scratches being quite

deep. Curve a gives the reflectivity of the same mirror after

giving it the final polish with graphite and alcohol, when only very

fine scratches remained. In the latter case the scattering is

almost eliminated, and the reflectivity rises abruptly from 39 per

cent in the yellow to 85 per cent at 2.5/x, beyond which point the

rise is gradual to 94 per cent at g/n. The theoretical reflecting

power, at 12/z, computed from the electrical conductivity, is about

95.7 per cent.

1000

90

80

%^ 70

g

h 60

50

40

,

7 a;y^PcT

i

I4> 5

TANTALUM10/a

Fig. 4

Through the kindness of Dr. Von Wartenberg an opportunity

was granted to examine the tantalum regulus mirror used by him

in determining the optical constants in the visible spectrum.

The area of surface which could be used was only about 2.5 mmx 3 mm, and hence it was impossible to obtain the absolute reflec-

tivities with high accuracy. There was no difficulty in obtaining

the relative values for different parts of the spectrum, which is

the point of most interest. The surface of the mirror was per-

fectly free from scratches. The reflectivity, which is shown in

210 Bulletin of the Bureau of Standards \Voi.7,No.2

curve b, Fig. 4, rises even more abruptly to high values than was

found in the first sample. In this respect, pure tantalum, zinc,

and silver stand unique among the metals. Because of the steep-

ness of the reflectivity curve, the emissivity of tantalum in the

visible spectrum must be more selective than obtains in tungsten

and molybdenum. Furthermore, since the reflectivity of pure

tantalum is higher than that of tungsten at 1 to 2/i, it must have

the higher luminous efficiency.

The reflectivity of the regulus of pure tantalum as observed bytwo methods (Wartenberg computed R from the "optical con-

stants") is about 45 per cent in the yellow. The other mirror

appeared much darker, so that the lower reflectivities as shownin Fig. 4 are to be attributed to impurities. The tantalum reg-

ulus appeared as bright as the prism of tungsten already described.

The most casual observation shows that the tungsten filament has

a much whiter metallic luster than a tantalum filament. This

may be due to contamination in drawing the tantalum wire,

especially if drawn with oil as a lubricant. It appears that in

tungsten all the impurities can be driven off at high temperatures.

This is not necessarily true of tantalum, which occupies an anom-

alous position in the scale of luminous efficiencies of metals.

From the observations of Waidner and Burgess, who found that

the "normal" operating temperature of tantalum is about 2000 C,

it is evident that its low luminous efficiency is not due to a low

operating temperature, such as would be necessary with platinum.

Incandescent lamps of tungsten and osmium have an efficiency

of about 1.25 watts per candle, while for no apparent reason tan-

talum must be classified with graphitized carbon with an efficiency

of about 2 to 2.5 watts per candle.

The foregoing data on the reflectivity of tantalum and graphite

(Fig. 3) shows that this classification is consistent. The radiation

constants of these two substances are almost the same (total

radiation; tantalum proportional to the 5.3-power of temp.;

graphitized filament proportional to the 5.1 -power of temp.), and

the manner in which they differ is in the right direction, since

graphite is a nonmetal. Hence, while it was somewhat contrary

to expectation to find such a low reflectivity in the infra-red for

impure tantalum, this is not inconsistent with other data such as

the radiation constants.

Cobientz] The Reflecting Power of Various Metals 211

There seems to be a prevailing notion that the polish of the

metal filament is of great importance, but it is not apparent howroughness or scratches can cause a "blackening" of the radiation

by successive reflection within the cavities so long as the latter

are of the order of the wave length emitted. It is the absorption

coefficient and the refractive index that must be considered. This

is especially conspicuous in tantalum filaments. When new they

have a brownish luster, while the filaments of old lamps are decid-

edly black. Similar filaments of osmium retain their gray appear-

ance even after becoming crystalline. Under the microscope the

crystalline tantalum filament appears much darker than tungsten,

although the actual polish of the crystal surfaces of the latter is

much higher. Some crystals appear as dark as highly polished

jet or fine grains of stibnite, which has a reflectivity of only about

35 per cent. This low reflectivity in the visible spectrum produces

a high emissivity, but, contrary to the properties of the pure

material, the reflectivity continues low (emissivity high) in the

infra-red, which tends to lower the luminous efficiency.

The ideal solid illuminant must fulfill the conditions of a high

operating temperature, and a high emissivity in the visible spec-

trum. As shown in Fig. 1, of all the metals thus far examined,

tungsten approaches nearest to the fulfillment of these conditions.

ANTIMONY

It was found impossible to produce a perfect mirror of antimony,

so that the absolute values obtained have little weight. Since

the main object in the present work is to show the relative values

of reflectivities in the visible as compared with the infra-red, the

results obtained on several antimony mirrors are here recorded.

As with all the other metals, larger surfaces, either plane or of

known curvature, must be used for high accuracy in the absolute

values.

Two samples of antimony were examined, the one being a

mirror (5 mm x 1 7 mm) polished upon a cleavage plane of a large

crystal of the metal (from Kahlbaum), the other being a mirror

formed in vacuo by cathode disintegration. The reflectivity of

the polished cleavage piece is shown in curve a of Fig. 5. The

reflectivity is low, due to the presence of numerous large-sized

212 Bulletin of the Bureau of Standards [Vol. 7, No 2

pores which resulted in the polishing. Two cathode mirrors were

also examined. The first one contained small holes caused by

dust particles on the glass surface, which seemed to permit radia-

tion from the rear surface of the glass (2 mm in thickness) to

return on its path, thus causing a wavy reflection curve with

maxima of about 65 per cent at 1.2/A, 2.5/*, and 4/<i, and minima of

55 to 60 per cent at 2/*, 3.3/*, and 6/jl, beyond which point the

reflectivity increased gradually to 68 per cent at 9/x. On removing

the rear mirror surface the maxima and minima were destroyed.

A second mirror (5 mm x 17 mm), free from dust holes, and with

the rear surface of glass free from metal, was examined, the data

being plotted in curve b, Fig. 5. The reflectivity rises from 53

per cent in the yellow to a fairly uniform value of 72 per cent at 9//.

5 6

ANTIMONY

Fig. 5

In view of the difficulty in producing a mirror which is free from

carbon, by the cathode discharge, and in view of the fact that the

mirror seemed darker than one would expect, judging from the

appearance of the crystals, it seems probable that the reflectivity

is somewhat higher than here recorded. However, in view of the

fact that antimony departs considerably from the true metals, the

low reflectivity throughout the spectrum is probably to be expected,

as will be noticed presently in several other metals which lie on

the border line between the electrical conductors and the insu-

lators.

In conclusion, it may be added that cathode mirrors of antimony

are easily produced in 5 to 7 minutes, by using a heavy current.

The tendency is for one to continue the discharge for a longer

period than this, with the resultant oxidization of the central part

Cobientz) The Reflecting Power of Various Metals 2 1

3

of the mirror. By using a residual atmosphere of hydrogen anda low current, excellent large-sized mirrors are produced in half

an hour.RHODIUM

The sample examined was kindly loaned by Dr. Von Warten-

berg. The area of the surface examined was 5 mm x 5 mm. Sev-

eral small blowholes were present, which, with the smallness of

the surface, did not permit high accuracy in the absolute values

of the reflectivities. The reflecting power (Fig. 6), Rh, rises

abruptly from 77 per cent in the yellow to 92 per cent at 2.5/£,

beyond which point the rise is gradual to about 94 per cent at 9/z.

IRIDIUM

The mirror examined was a thick plate of the metal, which in

the polishing had not been perfectly freed from pores, which

caused scattering. The reflectivity in the yellow was only 53

per cent, which increased to 75 per cent at 2/u, and to 90 per cent

at 8/x. This same plate was examined several years ago 8 whenit had a finer polish. The reflectivity, Fig. 6, Ir, was found to

increase from 79 per cent at 1// to 95 per cent at gp. Its reflec-

tivity seems to be somewhat higher than platinum in the visible

spectrum.IRON

The sample of iron examined had a purity of 99.8 per cent.

The chief impurities were .15 per cent copper, .02 per cent man-

ganese, and .02 per cent carbon. The sample took a high polish

and reminded one somewhat of palladium in its general appear-

ances. The reflectivity, Fig. 6, Fe, rises gradually with increase

in wave length and, throughout the spectrum, the values are

about 2 per cent higher than the values previously observed on

steel by Hagen and Rubens.

MAGNESIUM

The mirror examined had a surface of about 1.2 cm x 2.5 cm.

It was polished with considerable difficulty. A highly polished

surface free from scratches was finally obtained by wet-grinding

8 See this Bulletin, 2, p. 470, 1907, where the reflectivity data of this and numerous

other metals is tabulated.

214 Bulletin of the Bureau of Standards [Vol. 7, No. 2

on fine emery paper, using tin oxide ("putty powder") and then

(dry) polishing the surface on chamois skin with a little putty

powder. This procedure prevented the formation of a film of

oxide. However, the surface was not entirely free from blisters,

so that the reflectivities are somewhat higher than recorded in

Fig. 6, Mg. The reflectivity curve is somewhat different from

that of a pure metal, especially of aluminum. (See this Bulletin,

vol. 2, p. 470.) The alloy of aluminum and magnesium, mag-

nalium, has a reflectivity curve which falls between these two

metals.

2345678 9/A

RHODIUM, IRIDIUM, IRON, MAGNESIUM, CHROMIUM

Fig. 6

CHROMIUM

The sample of chromium was sawed from a lump of the metal

made by the Goldschmidt process. The material was quite

crystalline, but took a fairly high polish, excepting the usual

pores which were rather large as compared with those of the

sample of tungsten, just described. The reflection curve, Fig.

6, Cr, increases gradually from 55 per cent in the yellow to 91

per cent at 9/i. Judging from previous experience with other

metals, it seems quite certain that, for a perfect mirror surface of

Coblentz] The Reflecting Power of Various Metals 215

noncrystalline material, the reflectivity of chromium would be

higher in the yellow, as observed by Wartenberg. While part of

the low reflectivity of chromium is no doubt due to diffuse reflec-

tion, out to 5//, it is to be noted that this metal produces acid com-pounds, just as was observed with antimony, and in its general

properties can not be classed with the true metals. Hence its

unusual reflection spectrum is to be expected.

80£

•

70 *^-*'

CTIVITY

o^,HI_J

U 50OS

s-r-*"--r

40pb

» ~

301

*

5 ; 2 <[ :> ti 1

-i) I^

TELLURIUM

Fig. 7

VANADIUM

The sample examined was kindly loaned by Dr. Von Warten-

berg. The surface was plane, highly polished, and the area

examined was about 5 mm x 5 mm.The reflecting power, curve c, Fig. 8, rises gradually from 58

per cent in the yellow to 92 per cent at 9/x. In this respect it is

very similar to chromium and iron, Fig. 6. This is to be expected

since they fall in the same group in the Mendelejefl series.

TELLURIUM

A knowledge of the reflectivity of tellurium is of interest because

it occurs on the border line between metals and nonmetals and

has a high electrical resistance. It is usually found in a highly

crystalline state, and it is therefore difficult to polish a surface free

216 Bulletin of the Bureau of Standards [Vol 7, No. 2

from pores. The reflectivity curve of such a mirror is given in

curve b, Fig. 7. It is of interest only in showing the general

trend of the reflectivities in various parts of the spectrum. The

absolute values would be very much higher if a correction could

have been made for the loss by scattering caused by the roughness

of the surface.

.

The best mirror used in the present examination was produced

by the cathode discharge in hydrogen. It was found that unusu-

ally large-sized mirrors can be deposited in this manner, the time

required being about half an hour. Thin deposits of tellurium

transmit a reddish-brown light, which would indicate a low

reflection beyond the red or a minimum reflection in the red.

This is illustrated in curve a, Fig. 7, where at .8//. the reflectivity,

after repeated observations, was found to be slightly less than in the

yellow. While this might be due to the lack of knowledge of the

true value of the reflectivity in the visible spectrum of the silver

comparison mirror, the color of the transmitted light seems to

indicate such a minimum. The occurrence of a reflectivity

minimum so far in the red is unusual in the true metals. How-ever, the characteristics of tellurium are so different from the

usual ones of true metals, that such an unusual reflecting power

is probably to be expected.

The construction of a thermopile having large surfaces usually

results in a sluggish instrument because of its great heat capacity.

It has been suggested to deposit by cathode discharge two metals

(for example, Te-Cu) upon some material which can afterwards be

removed. These two films of metal are to overlap along one edge,

which forms the active junction. The thermoelectric power of

tellurium is very high; and in view of the ease with which a film

of this metal can be deposited, it might prove satisfactory as a

surface thermopile.

SILICON

A knowledge of the reflectivity of this element is of interest

because it belongs to the nonmetals or insulators. Two samples

were examined. The polishing was done on a fine grade of emery

Coblentz] The Reflecting Power of Various Metals 217

paper covered with a mixture of tin oxide and a little graphite.

The one sample, from Kahlbaum, curve b, Fig. 8, was quite homo-

geneous and took a very high polish, free from scratches. The

second sample from the Carborundum Company, curve a, Fig. 8,

was less homogeneous, its hardness was greater and it took a poor

polish. The crystals as they come from the furnace are of a bluish

color, as was also true of the highly polished remelted sample

here examined. This causes a high reflectivity in the visible

(7? = 34 p. c. at .5/-1; 30 p. c. at .75/*; 28 p. c. at i/jl) spectrum,

which decreases rapidly to a uniform value of 28 per cent beyond

i/jl in the infra-red. The lower value of 24 per cent in curve a is

> 40

g Si

O 30ul_iu.Ill

K 20

90^

Va

80Nj>

70

0^"^"

60 \& h Si

/*s^

a * <*

1

a

50

3 4 5 6 7

SILICON, VANADIUM

]Q/x

Fig. 8

due to the porosity and to the consequent diffuse reflection by

this sample of silicon. From an examination of these two sam-

ples it appears highly improbable that the absolute values of the

reflectivities, to 10/x in the infra-red, are much greater than 30

per cent. It was noticed in Fig. 3 that the reflectivity of graphite

continues to rise gradually with increase in wave length. The

carbide of these two elements, carborundum SiC, as well as the

oxide Si0 2 , have the most remarkable 9 reflection spectra yet

observed, with bands of selective reflection which are as high in

value as the reflectivity of pure metals.

9 This Bulletin, 2, p. 476; 1907.

218 Bulletin of the Bureau of Standards [V01.7.N0.2

SUMMARY

The reflecting power of various metals has been examined,

including tungsten, tantalum, and molybdenum. By comparing

these metals with silver, of which the reflectivity is known, the

absolute reflectivities have been determined.

Each individual substance required a special mode of polishing

in order to produce a smooth surface free from contamination.

These methods are indicated in the text.

It is shown that, in common with all the other pure metals pre-

viously examined, the present series has a low reflectivity in the

visible spectrum which increases rapidly to high values in the

infra-red. The reflectivity curves of tungsten and of molybdenumare so nearly alike that from a consideration of their optical prop-

erties there is no great choice in the use of these two metals in

incandescent lamps. Since molybdenum is tough and tungsten

is brittle (i. e., becomes brittle after being operated at high tem-

peratures), it remains to be seen whether other physical diffi-

culties can be overcome in the former metal to enable its introduc-

tion as an illuminant.

The reflectivity curves show that the high efficiencies of the

metal filament lamps is due to their low reflectivity (true absorp-

tion, and not diffuse reflection, is meant) in the visible spectrum

and to a high reflectivity in the infra-red.

A high operating temperature is also necessary for a high lumi-

nous efficiency in metal filament lamps. This is exemplified in

the difference in the efficiencies of platinum and tungsten lamps.

In tungsten the low reflectivity in the visible spectrum results

in an emissivity of almost 50 per cent, while in the infra-red the

emissivity is less than 10 per cent of an ideal radiator.

The ideal solid illuminant giving a high luminous efficiency

must fulfill the condition of a high operating temperature, a high

reflectivity (hence low emissivity) in the infra-red and a low

reflectivity (high emissivity) in the visible spectrum. The pure

metals, tungsten and molybdenum fulfill these conditions moreexactly than any other known metals, excepting tantalum, which

as a lamp filament, however, does not have the optical prop-

erties of the samples described in this paper.

CMentz) The Reflecting Power of Various Metals 219

-2000

-200° —100° 0° 100°" 200° 300°

THERMAL EMF. MOLYBDENUM COPPER

-2800

400° 500c

Fig. 9

71132 —n 6

220 Bulletin of the Bureau of Standards [Voi.7,No.2

Thus far it has been impossible to obtain a large-sized sample

of pure osmium, to determine its reflectivity.

In conclusion, the writer expresses his appreciation of the cour-

tesy of Dr. Von Wartenberg in submitting his samples of tan-

talum, rhodium, and vanadium for examination.

NOTE I

ON THE THERMOELECTRIC PROPERTIES OF MOLYBDENUM

Having the requisite apparatus and pure wire at hand, it

seemed of interest to determine and place on record the thermo-

electric behavior of molybdenum. This seemed especially true

since its physical properties do not warrant a separate and exten-

sive investigation; and yet this data is of importance for com-

parison with other metals.

The methods and instruments here employed were described

in a previous paper on the thermoelectric properties of tantalum

and tungsten. 10 The wire used (specific resistance given on a pre-

vious page) was .627 mm in diameter and about 25 m long. Cop-

per wires .2 mm in diameter were wound tightly around the ends

of the molybdenum wire and soldered with tin. For the melting

points of zinc and tin the junction was in a thin glass tube. Above6oo° the molybdenum oxidized so rapidly that no measurements

were made on its emf. For temperatures as high as 419(m. p. zinc) the metal shows no oxidation. The observations

were extended to — 78 °. 5 C, which temperature was obtained

with a thick solution of carbon dioxide snow and ether. Theresults are given in Fig. 9, which includes tantalum. Between

o° and ioo° these two metals have almost the same thermoelectric

power. It is to be noticed, however, that they are opposite in

polarity. In a copper-tantalum couple the direction of the cur-

rent is the same as in copper-constantan, while in copper-molyb-

denum the direction of the current is opposite to that of a copper-

constantan couple. Between o° and ioo° the thermoelectric

power of the molybdenum-copper couple is 4.4, between ioo° and

200 it is 6.6, and between 200 and 400 it is 9.3 microvolts per

degree. The thermoelectric power, between o° and ioo°, is

about one-tenth the thermoelectric power of copper-constantan.

,0 This Bulletin, 6, p. 107; 1909.

Cobientz] The Reflecting Power of Various Metals 221

NOTE n

PRESERVATION OF SILVER MIRRORS

Silver-on-glass mirrors and especially mirrors made by grinding

and polishing the metal itself become tarnished and turn brownvery rapidly when exposed to the air in the laboratory. Thequestion naturally presents itself whether or not the slight differ-

ence in reflectivity of molybdenum and tungsten, described on a

previous page, is caused by a systematic error due to a gradual dete-

rioration of the silver comparison mirror. The electrical conduc-

tivity indicates that the reflectivity of molybdenum should be

slightly higher than tungsten. Since the two metals were ex-

amined in rapid succession on the same day, the question of

deterioration of the silver mirror is negligible. This silver-on-

glass mirror was prepared just before the observations were made,

but no difference could be detected between it and mirrors nowalmost two years old. The latter mirrors, some of them 15 cmin diameter, also a mirror of solid silver, were prepared in the

earlier part of 1908 and were at once placed in a large desiccator

containing vessels of phosphorous pentoxide and potassium

hydroxide. When not in use the mirrors are kept in this manner.

A large mirror spectrometer has its optical parts similarly inclosed

and the mirrors appear as white as when made two years ago.

This simple device of keeping the mirrors in dry air seems to

preserve them indefinitely, in marked contrast with mirrors

standing in the open, when the surface tarnishes in a very short

time.NOTE III

COMMENTS ON RADIATION LAWS OF METALS

In a previous paper n the spectral radiation constant, "a" of

various metals was described. This "constant" was found to

vary with the temperature and theoretical reasons were given to

account for this variation. The data presented in the present

paper support the experimental data then obtained, but lead

to a different and a more plausible explanation of the cause of

the variation in "or" with temperature. If in the previous

experiments the equation representing total radiation had been

11 Cobientz, this Bulletin, 5, p. 339; 1909. See p. 367 for a discussion of the radia-

tion from platinum.

222 Bulletin of the Bureau of Standards [Voi.7,No.z

considered, S = <rTa~ 1

, then we would have had to consider the

constant "a-" as well as the constant "or." The constant "c2,"

in the equation jc2= a\maxT, also involves the temperature, as

well as other factors, so that "a" was the only constant that

could be investigated, as described in the aforesaid paper.

Assuming " a value of a = 6, the radiation from platinum

observed by Iyummer and Kurlbaum 12, shows a variation in the

constant of total radiation from o- = 8.24 to o-=n.i6 for an

increase in temperature of about 1275 degrees.

It seems apparent that "<r" can not be considered constant, for

it indicates the emissivity per unit area of a surface radiating under

specific conditions. Now if those conditions vary, as is true of

the metals in which the absorptivity (at any wave length) is a

function of the temperature, it would seem that the "<r" can not

remain constant, but must increase with rise in temperature.

The coefficient of emissivity, "<r," must therefore be rigidly

defined, just as is true of the resistivity.

The constant, a, in the Wien equation E = c1\"~ae~c

2KiT and as

tentatively assumed in the aforesaid computations, relates to a

specific region of the spectrum; and irrespective of the writer's

published data, experiments by others show that the emissivity

of metals at any wave length in the infra-red is affected by the

temperature, i. e., it has a temperature coefficient. But this is

not all. At low temperatures the maximum emission, Emax ,

falls in the region of 3 to 6 \x where the reflecting power at a given

temperature is practically constant and we have to consider only

the change in the radiation constant, " a," due to the temperature

coefficient of the reflectivity. At high temperatures the Emax

falls in the region of 1 .5 //. (see Fig. 1) , where at any given tempera-

ture the reflectivity changes rapidly with wave length; and as it

shifts toward the short wave lengths the Emax must increase by a

different law (from the Wien equation Emax =B Ta) from what

it did when at 3 to 6 /jl. In the region of 1.5/i the temperature

coefficient is small, but the absorption (100-reflectivity) varies

12 Lummer and Kurlbaum, Verh. Phys. Ges., Berlin, 17, p. 106; 1898. Theyassumed a value of a=5 and found a variation of <r=4.28 to <r= 19.64 for a temperature

change of about 1275 .

CMentz] The Reflecting Power of Various Metals 223

rapidly with wave length, while beyond 3 p the conditions are

exactly the reverse. In both cases the emissivity, Emax , in-

creases in a different manner than postulated by the simple

radiation formulas.

Now, from the Wien equation, EK = cl

\~ae~Ci,kT, the radiation

constant "or" is computed from the equation (Eq. 1)

log E - log Emax

(Eq-

l }log e - ^

maxlog e + log ^

max

A. A.

where E and Emax are the emissivities at the wave lengths A. and

A-maxj respectively. It will be noticed that the temperature of

the filament does not enter into this equation. From the most

recent work of Hagen and Rubens, (Phys. Zeit. 2, p. 139, 1910)

it appears that in the region extending from the visible to 3 fi

the temperature coefficient of emissivity is extremely small, and

the observed variations in "a" with temperature, as computed

from the above equation, must be due to some other cause than

a true temperature coefficient of emissivity. This cause is not

difficult to find if we turn to the reflectivity curves, Fig. 1. At

low temperatures, when Xmax falls in the region of 2 /jl, the Emax is

but slightly affected by the temperature coefficient of emissivity,

and since the reflecting power varies but slowly with the wave

length, the Emax will be closely that required by the Wien equation

(Emax =B Ta). At high temperatures when the Xmax falls in the

region of 1.2 /*, the Emax falls in the region where the reflectivity

changes rapidly with wave length, and consequently the value

of Emax must increase at a higher rate than required by the equa-

tion Emax =B Ta. The difference between log E - log Emax (Eq. 1

)

must therefore be less than would otherwise be the case, due to

this fact, and the decrease in "a" with temperature when com-

puted from the above equation must necessarily follow, as

observed.

From the reflectivity curves this seems a plausible explanation.

Hence, an accurate value of "a" cannot be computed from the

Wien equation as expressed in (Eq. 1). However, from the close

agreement between the values of" a" of platinum by this and by

224 Bulletin of the Bureau of Standards [Voi.fi No. z

more direct methods, it is evident that the errors resulting from

the use of Equation i are not serious. The aforesaid work has

shown in a fairly quantitative way that the constant "or" of

tungsten is much higher than that of platinum and of carbon.

The great suppression of the radiation of tungsten as compared

with carbon in the infra-red is another proof of this. The Wienspectral energy curves, using different values of "a," illustrate

this same behavior, as may be easily verified by plotting the data

given in Table I, this Bulletin, 5, p. 347.

It seems evident that any spectral radiation formula which can

be set up must contain factors which will take into consideration

not only the variation in emissivity with temperature, but also

the variation in the absorptivity, which is a function of the wavelength and the temperature. Whether the constant "or" really

varies 13 as found in my previous paper where the Wien equation

was tentatively assumed, or whether the constant "<x" varies, or

whether both vary, is not the all important question at this stage

of our knowledge. The most important thing for the present is

to get further experimental data 13 on total and spectral radiation,

using wide strips of such metals as platinum and tungsten, and

suitable means to measure the temperatures. In this manner it

is hoped to contribute more accurate data in the near future.

Summary.—For metals the physical facts are:

(1) An extraordinary variation in their reflectivity, from the

visible to about 2 /jl, beyond which point the reflectivity increases

but slowly with wave length.

(2) A small temperature coefficient of emissivity from the

visible to 3/z, beyond which point the temperature coefficient in-

creases rapidly with wave length.

(3) For the black body the position of maximum emission in

the spectral energy curve shifts, with the rise in temperature,

toward the short wave lengths by a given amount, because only

the temperature enters the problem and \maxT = const. Further-

more the height of the maximum emission increases by a fixed

amount Emax = BTa. The total radiation progresses as S = aTa~ 1

.

The constants "or" and "c2

"are concerned in the slope of the

13 This conclusion was evident in my previous paper. This Bulletin, 5, p. 379; 1909.

Cobieniz) The Reflecting Power of Various Metals 225

spectral energy curve, while "<r" defines the total emissivity per

unit area of the radiator. In the pure metals, the variation in

Emax and Xmax will be affected (by opposite amounts) by (1) and

(2), and as already mentioned in the text, these formulae, just

mentioned, must be modified to take account of these facts.

(4) The cause of the constancy of "a" as found in the visible

spectrum and of its variability as found in the infra-red, is prob-

ably due to the fact that in the very limited region of the visible

spectrum \if and X2are close together, and, in the computations,

they are taken the same at different temperatures. Moreover no

temperature coefficient of emissivity has yet been definitely

established in this region. On the other hand, in the infra-red,

\ and \max are far apart, and Xmax is different for different tem-

peratures. Furthermore, the \max is affected by a temperature

coefficient of emission, and by a rapid change in emissivity due

to a rapid variation in the reflecting power, especially in the

region of 1 to 1.5//.

Washington, May 28, 19 10.